Continuous medium neural network coupling landslide prediction calculation method and device and medium

Abstract

The invention relates to the technical field of landslide prediction, and discloses a continuous medium neural network coupling landslide prediction calculation method and device, and a medium. The method comprises the following steps: delimiting a parameter space; according to the parameter space, landslides under different parameter combinations are simulated through a continuous medium theory single-phase flow model; wherein the information of the landslide under different parameter combinations is determined according to the single-phase shallow water wave model, and the information comprises at least one of the flow velocity, the flow depth and the accumulation field; determining an optimal landslide parameter combination according to the deep neural network and supervised learning; and inputting the optimal parameter group to the single-phase shallow water wave model, and carrying out final landslide dynamic process numerical simulation. According to the technical scheme of the invention, aiming at the high-speed remote debris flow phenomenon in China, from the perspective of particulate matter flow dynamics, the motion characteristics of the debris flow phenomenon are researched, and the landslide dynamic process simulation is realized. The prediction capability of the landslide debris flow starting-movement-disaster-causing range can be widely improved, and the method has important scientific significance for disaster prevention and reduction in mountainous areas in China.

Description

Technical Field

[0001] The invention relates to the technical field of landslide prediction, and in particular to a continuous medium neural network coupled landslide prediction calculation method, device and medium. Background Art

[0002] Landslide debris flow is a widespread geological phenomenon. Landslide debris flow disasters are triggered by multiple external factors such as earthquakes, volcanic eruptions, glacier melting, landslides, rainstorms, etc., with impact, siltation and erosion as the main hazards, which can cause serious harm to human life and property safety. It is a multiphase fluid composed of a large number of solid particles, a small amount of water medium and gas; driven by gravity, it can flow rapidly in valleys or on slopes. Its flow state is often complex and changeable, and most of them have the characteristics of high speed and long distance. The interaction between solid particulate matter in the debris flow is very complex, which comprehensively affects the fluidity of the debris flow.

[0003] Different from the rigid-body motion of general landslides, common debris flows move in a fluid-like manner and have obvious rheological properties. Debris flows are composed of a large number of loose particles, which are large in size and contact each other to form a skeleton. Contact forces play a major role in movement. The overall flow of debris flows often has the characteristics of high speed and long distance. Based on this, the ability to simulate and predict landslides through landslide dynamics needs to be improved. Summary of the invention

[0004] The present invention provides a continuous medium neural network coupled landslide prediction calculation method, device and medium to solve the above-mentioned existing technical problems.

[0005] According to a first aspect of the present invention, a continuum neural network coupled landslide prediction calculation method is provided, comprising:

[0006] Delimiting parameter space;

[0007] According to the parameter space, the landslide under different parameter combinations is simulated by a single-phase flow model of continuous medium theory; wherein, according to a single-phase shallow water wave model, information of the landslide under different parameter combinations is determined, and the information includes at least one of flow velocity, flow depth and accumulation field;

[0008] Determine the optimal landslide parameter combination based on deep neural network and supervised learning;

[0009] The optimal parameter group is input into the single-phase shallow water wave model to perform the final numerical simulation of the landslide dynamic process.

[0010] Preferably, in simulating the landslide under different parameter combinations by using the single-phase flow model of the continuum theory,

[0011] The computational model of landslide-debris flow blocking fluid mechanics is based on the assumption that the landslide-debris flow material is a simplified, three-dimensional, single-phase, isothermal solid phase material, and is solved by partial differential equations, namely the following formula:

[0012]

[0013]

[0014] In the formula, represents the partial derivative of density ρ with respect to time t, is the partial differential operator, ρ s is the solid density, is the divergence operator, is the vector cross product operator, u s 、v s is the velocity component of the landslide debris flow along the x and y directions, g is the gravitational acceleration, and T is the second-order stress tensor.

[0015] Preferably, solving U s The spatial and temporal distribution of the landslide-debris flow is obtained by simplifying the landslide-debris flow single-phase medium model through the depth integral average. The vertical distribution of the landslide-debris flow characteristics is ignored, so the planar distribution of the landslide-debris flow characteristics in space is solved, that is, the following formula:

[0016]

[0017] In the formula, h s is the depth of landslide debris flow, u s 、v s is the velocity component of the landslide debris flow along the x and y directions, f bx 、f by is the base friction of the landslide debris flow along the x and y directions, and the following formula for the fluid mechanics calculation model is available:

[0018]

[0019] Where U w is the characteristic velocity at the beginning of velocity weakening, ‖U‖ is the modulus of the velocity vector, μ w is the static friction coefficient, μ0 is the dynamic friction coefficient, g is the acceleration of gravity, h is the depth of the landslide, and θ is the slope foot.

[0020] Preferably, there is a mold resistance model Therefore, the relationship between formula 6 and μ(U) is:

[0021]

[0022] where μ(U) is the coefficient of friction as a function of the characteristic velocity.

[0023] Preferably, in determining the optimal landslide parameter combination based on deep neural network and supervised learning,

[0024] According to the deep neural network, supervised learning and verification are performed on the accumulation information under different simulated parameter groups and the situations under real conditions, and the landslide parameter group that best matches the real situation is determined through the deep neural network model, thereby obtaining the optimal landslide parameter combination.

[0025] Preferably, the deep neural network model is set as a BP neural network model, and the BP neural network model iteratively trains the network based on the propagation criterion of error gradient descent; the structure of the BP neural network model includes: an input layer, a hidden layer and an output layer; the input vector is: x = (x1, x2, ..., x i , …, x j ) T , the weight from the input layer to the hidden layer is represented by V, V = (v1, v2, ..., v i , …, v k ) T , where v i is the weight vector corresponding to the i-th neuron in the hidden layer; the output vector of the hidden layer is: Y = (y1, y2, ..., y i , …, v k ) T , the weight matrix from the hidden layer to the output layer is represented by W, W = (w1, w2, ..., w i ,…,w k ) T , where the vector w i is the weight vector corresponding to the i-th neuron in the output layer, and T represents the transpose of the vector.

[0026] Preferably, the BP neural network model training includes:

[0027] The input value is calculated by the neural network model to obtain an output, which is the predicted value, where each input corresponds to a known output;

[0028] Compare the difference between the predicted value and the known actual value, and then calculate the gradient value of the loss function in the network structure;

[0029] Modify the weights of each layer according to the feedback to minimize the loss function.

[0030] According to a second aspect of the present invention, a continuum neural network coupled landslide prediction and calculation device is provided, comprising:

[0031] A delineation module, used to delineate parameter space;

[0032] A simulation module is used to simulate the landslide under different parameter combinations through a single-phase flow model of continuous medium theory according to the parameter space; wherein, according to a single-phase shallow water wave model, the information of the landslide under different parameter combinations is determined, and the information includes at least one of flow velocity, flow depth and accumulation field;

[0033] A determination module for determining an optimal landslide parameter combination based on a deep neural network and supervised learning; and

[0034] The final simulation module is used to input the optimal parameter group into the single-phase shallow water wave model to perform the final numerical simulation of the landslide dynamic process.

[0035] According to a third aspect of the present invention, there is provided an electronic device, comprising the following steps:

[0036] Memory; and

[0037] processor;

[0038] Wherein, the memory is used to store one or more computer instructions; the one or more computer instructions are executed by the processor to implement any of the methods described above.

[0039] According to a fourth aspect of the present invention, a readable storage medium is provided, on which computer instructions are stored; wherein when the computer instructions are executed by a processor, any of the methods described above is implemented.

[0040] The technical solution of the present invention is aimed at the phenomenon of high-speed and long-distance debris flow in my country, starting from the perspective of particle material flow dynamics, studying its motion characteristics and realizing the simulation of landslide dynamic process. It can widely improve the prediction ability of the initiation, movement and disaster-causing range of landslide debris flow, and has important scientific significance for disaster prevention and mitigation in mountainous areas of my country. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of a flow chart of a continuous medium neural network coupled landslide prediction calculation method in an embodiment;

[0042] Figure 2 is a schematic diagram of the structure of a BP neural network in an embodiment;

[0043] Figure 3 is a schematic diagram of the structure of a landslide prediction model combining shallow water waves and neural networks in an embodiment;

[0044] Figure 4 is a picture of the location of the Baige landslide in October 2018 and the lateral views of two Baige landslides in one embodiment;

[0045] Figure 5is a numerical simulation result of Baige landslide coupling model in one embodiment;

[0046] Figure 6 The diagram is a schematic diagram of a continuous medium neural network coupled landslide prediction calculation device in an embodiment. DETAILED DESCRIPTION

[0047] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0048] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solution in one embodiment will be clearly and completely described below in conjunction with the drawings in one embodiment. Obviously, the described embodiment is only an embodiment of a part of the present invention, not all embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.

[0049] It should be noted that the terms "first", "second", etc. in the present invention are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present invention described herein. In addition, the terms "including the following steps" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed in the following steps or are inherent to these processes, methods, products or devices.

[0050] It should be understood that when an element (such as a layer, film, region, or substrate) is described as being "on" another element, the element may be directly on the other element, or there may be an intermediate element. Moreover, in the present invention, when an element is described as being "connected" to another element, the element may be "directly connected" to the other element, or "connected" to the other element through a third element.

[0051] Embodiment 1

[0052] Please refer to Figure 1 This embodiment provides a continuous medium neural network coupled landslide prediction calculation method, including the following steps:

[0053] S1. Define the parameter space; the parameter space may be a multi-dimensional parameter space;

[0054] S2. According to the parameter space, simulating the landslide under different parameter combinations by using a single-phase flow model of continuum theory; wherein, according to a single-phase shallow water wave model, determining the information of the landslide under different parameter combinations, the information including at least one of flow velocity, flow depth and accumulation field;

[0055] S3. Determine an optimal landslide parameter combination according to a deep neural network and supervised learning; wherein the deep neural network involves a deep neural network model, and the deep neural network model has been fully learned / trained;

[0056] S4. Input the optimal parameter group into the single-phase shallow water wave model to perform the final numerical simulation of the landslide dynamic process.

[0057] Wherein, in step S2, the single-phase shallow water wave model is based on the single-phase flow model. In one embodiment, the single-phase shallow water wave model can be understood as / is a single-phase flow model.

[0058] In one embodiment, the parameters that need to be manually set in the continuum simulation (i.e., the continuum theory single-phase flow model simulation in step S2) are the parameters in the friction model, referring to (friction) formula (6), specifically: U w is the characteristic velocity at the beginning of velocity weakening, μ w is the static friction coefficient, μ0 is the dynamic friction coefficient; however, there is often no basis for artificially set values, but it is known that each parameter has an approximate range, such as μ w , μ0 is a parameter value (0,1), U w is a parameter value of (0,10), so a parameter space μ with three parameters is drawn w (0,U w (0,10), and take the points of the parameter group equidistantly in the parameter space as the parameter values ​​of the simulation, such as (μ w ,μ0,U w )=(0.1,0.3,1)(0.2,0.3,1)……

[0059] In one embodiment, in step S2, the calculation model of the landslide-debris flow blocking fluid mechanics is solved by partial differential equations based on the assumption that the landslide-debris flow material is a simplified, three-dimensional, single-phase, isothermal solid-phase material, that is, the following formula:

[0060]

[0061] In the formula, represents the partial derivative of density ρ with respect to time t, is the partial differential operator, ρ s is the solid density, is the divergence operator, is the vector cross product operator, u s 、v s is the velocity component of the landslide debris flow along the x and y directions, g is the gravitational acceleration, and T is the second-order stress tensor.

[0062] Based on the above partial differential equation of fluid mechanics, solve U s The spatial and temporal distribution of the landslide-debris flow is simplified by the depth-integrated average to obtain the control equation of the single-phase medium model of the landslide-debris flow. The vertical distribution of the landslide-debris flow characteristics is ignored, so the plane distribution of the landslide-debris flow characteristics in space is solved, that is, the following formula:

[0063]

[0064] In the formula, h s is the depth of landslide debris flow, u s 、v s is the velocity component of the landslide debris flow along the x and y directions, f bx 、f by is the base friction of the landslide debris flow along the x and y directions, and the following formula for the fluid mechanics calculation model is available:

[0065]

[0066] Among them, the modulus resistance model μ(U) is a part of Formula 6, Therefore, the relationship between formula 6 and μ(U) is:

[0067]

[0068] Where μ(U) is the friction coefficient as a function of the characteristic velocity;

[0069] Where U w is the characteristic velocity at the beginning of velocity weakening, ‖U‖ is the modulus of the velocity vector, μ w is the static friction coefficient, μ0 is the dynamic friction coefficient, g is the acceleration of gravity, h is the depth of the landslide, and θ is the slope foot.

[0070] In one embodiment, in step S3, supervised learning and verification are performed on the accumulation information under different simulated parameter groups and the conditions under real conditions according to the deep neural network, and the landslide parameter group that best matches the real situation is determined through the deep neural network model to obtain the optimal landslide parameter combination.

[0071] In one embodiment, the deep neural network model is set to a BP neural network model, which is the most widely used ANN network model. The BP neural network is a multi-layer feedforward network trained using the BP algorithm, and is a supervised learning algorithm based on back propagation. The BP neural network model iteratively trains the network based on the propagation criterion of error gradient descent; the structure of the BP neural network model includes: an input layer, a hidden layer, and an output layer; the input vector is: X = (x1, x2, ..., x i , …, x j ) T , the weight from the input layer to the hidden layer is represented by V, V = (v1, v2, ..., v i , …, v k ) T , where v i is the weight vector corresponding to the i-th neuron in the hidden layer; the output vector of the hidden layer is: Y = (y1, y2, ..., y i , …, v k ) T , the weight matrix from the hidden layer to the output layer is represented by W, W = (w1, w2, ..., w i ,…,w k ) T , where the vector w i is the weight vector corresponding to the i-th neuron in the output layer, and T represents the transpose of the vector.

[0072] In one embodiment, the BP neural network model training includes the following steps:

[0073] S31. The input value is calculated by the neural network model to obtain an output, which is the predicted value, wherein each input corresponds to a known output;

[0074] S32. Compare the difference between the predicted value and the known actual value, and then calculate the gradient value of the loss function in the network structure;

[0075] S33. Modify the weights of each layer according to the feedback to minimize the loss function.

[0076] The continuous medium deep neural network coupled landslide prediction numerical calculation method of the present invention first defines the parameter space, and uses the single-phase shallow water wave model to determine the flow velocity, flow depth, and accumulation field information of the landslide under different parameter combinations. Through a type of deep neural network, the accumulation information under different simulated parameter groups and the situation under real conditions are fully supervised and verified. The optimal landslide parameter group that best matches the actual situation is determined through the fully learned deep neural network model; on this basis, the determined optimal parameter group is input into the single-phase shallow water wave model to obtain the final landslide dynamic process simulation.

[0077] The method of the present invention combines the numerical simulation of fluid dynamics of continuous medium theory and the deep learning neural network model. The calculation process is direct and the calculation results are highly accurate, providing a new idea of ​​cross-disciplinary coupling model for the prediction, prevention, preview and pre-treatment of landslide debris flow.

[0078] Embodiment 2

[0079] Please refer to Figure 1-Figure 5 This embodiment provides a continuous medium deep neural network coupled landslide prediction calculation method, which is explained from the perspective of research and development or technical implementation to facilitate reading and understanding; in order to clearly and completely explain the technical solution of the present invention and highlight the inventive points and key points of the technical solution of the present invention.

[0080] The continuous medium deep neural network coupled landslide prediction calculation method relates to a shallow water wave neural network coupled landslide prediction method. The specific operation steps of the shallow water wave neural network coupled landslide prediction method are as follows: first, a landslide debris flow database of various scales is established worldwide through data collection, and then the source location and source quantity of each data landslide are obtained through the established database, and the landslide dynamic process is predicted based on the fluid mechanics calculation model (formula (6)) and the single-phase shallow water wave model (formula (1)-(5)). Then, the parameters are screened through the deep learning neural network model to improve the prediction accuracy of the landslide debris flow.

[0081] Among them, it involves the calculation method of fluid mechanics for landslide-debris flow blocking the river. Its calculation model is based on the assumption that the landslide-debris flow material is a simplified, three-dimensional, single-phase, isothermal solid phase material, and is solved by partial differential equations, namely the following formula:

[0082]

[0083] The above partial differential equation of fluid mechanics, solve U s The spatial and temporal distribution of the landslide-debris flow is simplified by depth-integrated averaging to obtain the control equation of the single-phase medium model. The vertical distribution of the landslide-debris flow characteristics can be ignored, so the plane distribution of its characteristics in space can be solved, that is, the following formula:

[0084]

[0085] In the formula, h s is the depth of landslide debris flow, u s 、v s is the velocity component of the landslide debris flow along the x and y directions, f bx 、f by is the base friction of the landslide debris flow along the x and y directions, and the following formula for the fluid mechanics calculation model is available:

[0086]

[0087] In the friction model μ(U) formula, μ(U) is a function of the friction coefficient as a characteristic velocity, that is:

[0088] The relationship between formula 6 and μ(U) can be obtained as follows:

[0089]

[0090] Where U w is the characteristic velocity at the beginning of velocity weakening, ‖U‖ is the modulus of the velocity vector, μ w is the static friction coefficient, μ0 is the dynamic friction coefficient.

[0091] Among them, the values ​​of friction resistance model parameters are shown in the following table.

[0092] Table 1 Friction model parameter values

[0093]

[0094] In one embodiment of a deep learning neural network model, the neural network ANN is a non-parametric intelligent model, which is an intelligent system based on the understanding of the physiological structure of human brain tissue and its operating mechanism. The system can solve complex nonlinear mapping problems, and is particularly advantageous for complex mathematical problems that are difficult to solve by establishing accurate mathematical models. The model structure has the characteristics of high intelligence, strong robustness, and insensitivity to local abnormal point disturbances, and is widely used in the analysis of various types of data. According to the connection mode between the levels, the neural network can be divided into a feedforward neural network and a feedback neural network. Among them, the BP neural network is the most widely used ANN network model at present. The network is a multi-layer feedforward network trained using the BP algorithm, which is a supervised learning algorithm based on back propagation. The network structure iteratively trains the network based on the propagation criterion of error gradient descent, and has strong generalization ability and nonlinear mapping ability. Figure 2 A typical BP neural network structure is shown. The network consists of three layers: input layer, hidden layer and output layer. The input vector is: X = (x1, x2, ..., x i , …, x j ) T , the weight from the input layer to the hidden layer is represented by V, V = (v1, v2, ..., v i , …, v k ) T , where v i is the weight vector corresponding to the i-th neuron in the hidden layer; the output vector of the hidden layer is: Y = (y1, y2, ..., y i, …, v k ) T , the weight matrix from the hidden layer to the output layer is represented by W, W = (w1, w2, ..., w i ,…,w k ) T , where the vector w i is the weight vector corresponding to the i-th neuron in the output layer.

[0095] Furthermore, when training based on the BP neural network, it is required that each input has a known output. The input value can obtain an output through the calculation of the neural network model, which is the predicted value; the difference between the predicted value and the known actual value is compared, and then the gradient value of the loss function in the network structure is calculated, and finally the weights of each layer are modified and updated to minimize the loss function.

[0096] In general, a continuous medium neural network coupled landslide prediction calculation model, namely, a shallow water wave neural network coupled landslide prediction model, is formed based on the above methods. Figure 3 The final landslide prediction model is a combination of shallow water waves and neural networks. The basic process includes: first, using our single-phase shallow water wave model to generate a parameter space, and then performing numerical simulation according to the parameter combination in the parameter space. Then, the collected simulated landslide accumulation fan results and real sedimentary fan information are input into the deep neural network for training. In this way, the optimal parameter set of the deep neural network model can be finally obtained, and the parameter set is applied again to the single-phase shallow water wave model for the prediction of the final landslide range.

[0097] A case study is provided for the continuous medium neural network coupled landslide prediction calculation method of the present invention: the Baige landslide in Jinsha River in 2018.

[0098] Figure 4 In the figure, (a) shows the location of the Baige landslide in 2018; (b) shows lateral photos of the two Baige landslides that occurred on October 10 and November 3, 2018 (Photo source: Yangtze River Water Resources Commission).

[0099] Numerical simulation results of Baige landslide depth average equation in October 2018. Firstly, the traditional depth average model was used to numerically simulate the Baige landslide event in October 2018. The friction parameters used in the model are shown in the table above. The initial volume of Baige landslide is set to 2.5×10 7 m 3, and the landslide initiation mechanism is not considered in the simulation, and the landslide body starts to move at t=0s. The landslide body starts on the slope under the action of gravity and slides downward. After starting, the flowing mass gains momentum when moving downward. When t=40 seconds, the landslide slides into a steeper area, and the speed begins to increase sharply due to the smaller foundation friction and the larger slope. The landslide slides into the Jinsha River and gradually stops. However, when the traditional depth average model is used to simulate the Baige landslide, the accumulation results (accumulation range) under real conditions are quite different.

[0100] Figure 5 In the figure, (a)-(f) show the numerical simulation results of the Baige landslide coupling model of the present invention. The Baige landslide event on October 10, 2018 was numerically simulated under the same initial conditions using the technical solution proposed by the present invention. The initial volume of the Baige landslide is consistent with the traditional model, which is 2.5×10 7 m 3 The landslide initiation mechanism is not considered in the simulation. The landslide body starts to move at t = 0s and starts to slide down the slope under the action of gravity. Figure 5 In (d)-(f), the landslide slid into the Jinsha River and gradually stopped. The results of the coupled model proposed in the present invention in simulating the accumulation fan of the Baige landslide are basically consistent with the shape of the accumulation fan in the real situation, which is a significant improvement compared with the traditional model.

[0101] The continuous medium neural network coupled landslide prediction calculation method of the present invention, based on field prototype research, studies the rheological characteristics of debris flow through a combination of theoretical analysis, continuous medium mechanics model (deep integral equation) and deep learning (neural network model), and proposes a model of debris flow shear rate on bed (base) friction resistance. Then, the deep learning neural network model is used to perform high-dimensional parameter screening on the existing friction resistance model, thereby further improving the ability to predict landslide debris flow.

[0102] Embodiment 3

[0103] Please refer to Figure 6 This embodiment provides a continuous medium neural network coupled landslide prediction and calculation device, which adopts the following form:

[0104] 1. Delineate modules

[0105] The delimiting module 10 is used to delimit the parameter space;

[0106] 2. Analog module

[0107] The simulation module 20 is used to simulate the landslide under different parameter combinations through the single-phase flow model of the continuum theory according to the parameter space; wherein, according to the single-phase shallow water wave model, the information of the landslide under different parameter combinations is determined, and the information includes at least one of the flow velocity, flow depth and accumulation field;

[0108] 3. Determine the module

[0109] The determination module 30 is used to determine the optimal landslide parameter combination based on the deep neural network and supervised learning;

[0110] 4. Final simulation module

[0111] The final simulation module 40 is used to input the optimal parameter group into the single-phase shallow water wave model to perform the final numerical simulation of the landslide dynamic process.

[0112] It should be noted that the above device is used to implement the methods in the above embodiments, and each module in the device corresponds to each step in the method.

[0113] Embodiment 4

[0114] Based on the same inventive concept, an embodiment of the present invention provides an electronic device, comprising the following steps: a memory and a processor; wherein the memory is used to store one or more computer instructions; and the one or more computer instructions are executed by the processor as a method of any one of the above embodiments.

[0115] Embodiment 5

[0116] Based on the same inventive concept, an embodiment of the present invention provides a readable storage medium, on which computer instructions are stored; wherein, when the computer instructions are executed by a processor, any method in the above embodiments is implemented.

[0117] Among them, the above one or more computer instructions can be formed into a program.

[0118] The above program can be run in a processor, or it can be stored in a memory (or computer-readable medium), which includes the following steps: Permanent and non-permanent, removable and non-removable media can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, modules of programs or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer-readable media does not include the following steps: temporary storage of computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0119] These computer programs can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps of the functions specified in one or more blocks can be implemented by different modules corresponding to different steps.

[0120] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A continuous medium neural network coupled landslide prediction calculation method, characterized in that: include: Delimiting parameter space; According to the parameter space, the landslide under different parameter combinations is simulated by a single-phase flow model of continuous medium theory; wherein, according to a single-phase shallow water wave model, information of the landslide under different parameter combinations is determined, and the information includes at least one of flow velocity, flow depth and accumulation field; Determine the optimal landslide parameter combination based on deep neural network and supervised learning; The optimal parameter group is input into the single-phase shallow water wave model to perform the final numerical simulation of the landslide dynamic process.

2. The continuous medium neural network coupled landslide prediction calculation method according to claim 1 is characterized in that: In the simulation of landslides under different parameter combinations using the single-phase flow model of continuum theory, The computational model of landslide-debris flow blocking fluid mechanics is based on the assumption that the landslide-debris flow material is a simplified, three-dimensional, single-phase, isothermal solid phase material, and is solved by partial differential equations, namely the following formula: In the formula, represents the partial derivative of density ρ with respect to time t, is the partial differential operator, ρ s is the solid density, is the divergence operator, is the vector cross product operator, u s 、v s is the velocity component of the landslide debris flow along the x and y directions, g is the gravitational acceleration, and T is the second-order stress tensor.

3. The continuous medium neural network coupled landslide prediction calculation method according to claim 2 is characterized in that: Solving for U s The spatial and temporal distribution of the landslide-debris flow is obtained by simplifying the landslide-debris flow single-phase medium model through the depth integral average. The vertical distribution of the landslide-debris flow characteristics is ignored, so the planar distribution of the landslide-debris flow characteristics in space is solved, that is, the following formula: In the formula, h s is the depth of landslide debris flow, u s 、v s is the velocity component of the landslide debris flow along the x and y directions, f bx 、f by is the base friction of the landslide debris flow along the x and y directions, and the following formula for the fluid mechanics calculation model is available: Where U w is the characteristic velocity at the beginning of velocity weakening, ‖U‖ is the modulus of the velocity vector, μ w is the static friction coefficient, μ0 is the dynamic friction coefficient, g is the acceleration of gravity, h is the depth of the landslide, and θ is the slope foot.

4. The continuous medium neural network coupled landslide prediction calculation method according to claim 3 is characterized in that: Modular resistance model Therefore, the relationship between formula 6 and μ(U) is: where μ(U) is the coefficient of friction as a function of the characteristic velocity.

5. The continuous medium neural network coupled landslide prediction calculation method according to claim 1 is characterized in that: In the determination of the optimal landslide parameter combination based on deep neural network and supervised learning, According to the deep neural network, supervised learning and verification are performed on the accumulation information under different simulated parameter groups and the situations under real conditions, and the landslide parameter group that best matches the real situation is determined through the deep neural network model, thereby obtaining the optimal landslide parameter combination.

6. The continuous medium neural network coupled landslide prediction calculation method according to claim 5 is characterized in that: The deep neural network model is set as a BP neural network model, and the BP neural network model iteratively trains the network based on the propagation criterion of error gradient descent; the structure of the BP neural network model includes: an input layer, a hidden layer and an output layer; the input vector is: X = (x1, x2, ..., x i , …, x j ) T , the weight from the input layer to the hidden layer is represented by V, V = (v1, v2, ..., v i , …, v k ) T , where v i is the weight vector corresponding to the i-th neuron in the hidden layer; the output vector of the hidden layer is: Y = (y1, y2, ..., y i , …, v k ) T , the weight matrix from the hidden layer to the output layer is represented by W, W = (w1, w2, ..., w i ,…,w k ) T , where the vector w i is the weight vector corresponding to the i-th neuron in the output layer, and T represents the transpose of the vector.

7. The continuous medium neural network coupled landslide prediction calculation method according to claim 6 is characterized in that: BP neural network model training includes: The input value is calculated by the neural network model to obtain an output, which is the predicted value, where each input corresponds to a known output; Compare the difference between the predicted value and the known actual value, and then calculate the gradient value of the loss function in the network structure; Modify the weights of each layer according to the feedback to minimize the loss function.

8. A continuous medium neural network coupled landslide prediction and calculation device, characterized in that: include: A delineation module, used to delineate parameter space; A simulation module is used to simulate the landslide under different parameter combinations through a single-phase flow model of continuous medium theory according to the parameter space; wherein, according to a single-phase shallow water wave model, the information of the landslide under different parameter combinations is determined, and the information includes at least one of flow velocity, flow depth and accumulation field; A determination module for determining an optimal landslide parameter combination based on a deep neural network and supervised learning; and The final simulation module is used to input the optimal parameter group into the single-phase shallow water wave model to perform the final numerical simulation of the landslide dynamic process.

9. An electronic device, characterized in that: The following steps are involved: Memory; and processor; Wherein, the memory is used to store one or more computer instructions; the one or more computer instructions are executed by the processor to implement the method described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores computer instructions; wherein, when the computer instructions are executed by a processor, the method described in any one of claims 1 to 7 is implemented.

Images